ABSTRACT
Faith-based provision and reception of social services may yield results that differ either randomly or persistently (via “fixed effects”) from the outcomes achieved under non-faith-based, but certainly not value-free, administration of these services. Yet attributing raw differences in performance to unobservable factors assumed to derive from “faith” should be a last resort. Instead, the objective of analysis should be to identify any such factors and their cost and strength operationally so that they can become a managed part of any program offered by existing providers or entrants.
These factors, or the cost effectiveness of employing them, could differ between secular and faith-based providers. Then members of the two groups optimally would use a different mix of means even if they were to pursue precisely the same ends. Yet program design and operation by both groups would stand to benefit from knowing the keys to their relative performance in particular areas of social service with particular groups of clients. Comparing the outcomes of experiments based on 40random assignment may provide little help in that regard. Micro-simulations may provide a superior approach to tightening the link between input-based model predictions and continuous learning from outcomes.
